I've just finished reading On Intelligence by Jeff Hawkins - the father of the PalmPilot and founder of both Palm and Handspring. I remember reading an interview with Jeff a few years back where he talked about how, despite his involvement in mobile, what he really wanted to do was get some of his theories about the nature of intelligence down on paper - and he's done a really good job.

In the book, he posits that building artificially intelligent machines is eminently possible - that there's nothing magical about intelligence itself that precludes this. In fact he sums this up beautifully in a chapter on consciousness, stating that "consciousness is just what it feels like to have a neocortex". He then lays out his theories for how the brain works: that basically, our minds use hierarchical structures of layers of neurons to constantly make predictions about the world our senses show to us, and feeding these predictions back into the input from our senses to evaluate them.

It's an interesting theory, and one I'm looking forward to discussing with the FP crew, 3 of whom came out of the Cognitive Science department of Sussex University and know much more about this stuff I. And I was all fired up by the epilogue, an enthusiastic call for students interested in the topic to get working on it - after decades wandering about the wilderness, Jeff is convinced that we're due for a burst of progress in this area in the next 10 years.

Thoroughly recommended - and there's an associated web site, of course.

Disclaimer: I'm a frustrated AI wannabe. Despite living in Brighton I avoided Sussex University, which has a great rep for such things, on the basis that I didn't want to spend my university years living in the family home... and I failed to get into Edinburgh, the other big UK AI centre, thanks to my dreadful maths.

Comments

[Dave Ph]

Hmmm.

I've not read the book. In fact I've not read anything on AI in years.

But I have read every issue of New Scientist from cover to cover for the last four years. And I'm sure I would have remembered if somewhere in there there had been a 'Basis of Consciousness Discovered!' article. Because I'm thinking that to develop Artificial Intelligence, you kind of need to know haw 'natural' intelligence works.

Otherwise, how would you know you have created it?

Maybe I'm doing Hawkins a disservice. But I'm scared by the amazing amount of pop-sci 'how your brain works' which swiftly leads to 'and then you scan in the state of each neuron and upload it to the computer - you don't even notice the transition...' type transhumanist b****x. I gather Hawkins doesn't mention the singularity. Maybe he's on the side of good after all...

[Tom Hume]

One of the things he covers in some depth in his book is some theories on how the brain works - which are what he's looking to model.

He doesn't claim to have all the answers, and at one point declares that some of what he's put down is certainly wrong - but it's a theory I'd not come across before, and there are a few testable predictions in a later chapter, which he offers to prove or disprove his theories.

Want to borrow it? It'll at least mean you can reject his theories from a slightly firmer foundation than "I didn't see it in The New Scientist so it can't be right" - and I think I owe you a book ;)

In the meantime I'm going to shove this down the throats of someone who's had recent contact with the field; I'll post here when I get some feedback from them...

[Dave Ph]

I'd love to read it, but I now have over a metre of books on my pile... Maybe one day next year :-(

Cheers